Forward-looking sensors provide better accuracy, but that's only beneficial to military airplanes that rely on terrain-following for deep penetration missions. Civil planes don't need such accuracy because they shouldn't get that close to terrain.
And while yes line-of-sight is typically a none issue, this only applies to what's ahead. That radar can't see what's say 90° to the right when a turn is executed to the right; an EGPWS database will already paint the navigation display with such terrain.
Both systems can be blended, but for civil planes, it's an extra cost that is not needed because getting that close to terrain is out of the question. There are also export restrictions with terrain sensing systems – which for example shelved a synthetic-aperture radar landing system that had interest from civil operators in the mid-90s, because an advanced system that is designed to ignore wrong reflections from buildings and vegetation, could be adapted into an adversary's air force (planes and cruise missiles).
➤ Assuming a basic system to avoid export-embargoes, and a low-gain to avoid false-positives, then it will not be better than a database-based system (and will be worse for terrain that surrounds a plane; the big turn example). That's the same issue with radar-altimeter (RA) GPWS:
These systems [meaning RA-based] are also subject to high false alarm rates, particularly when landing approaches occur over rugged terrain. High false alarm rates hinder pilot acceptance and trust of such systems and can cause valid warnings to go unheeded.
All those options were looked into. Here's a visual summary of the systems from one of those studies, with the forward-looking element being optional:
Image and quotation source: Young, Craig S. "Warning system concepts to prevent controlled flight into terrain (CFIT)." 1993.